Geri Dön

Esnek üstyapılarda makine öğrenmesi yöntemleri ile pürüzlülük tahmini

Roughness estimation in flexible pavements using machine learning methods

PDF İndir

  1. Tez No: 924501
  2. Yazar: HÜSEYİN ÇUHA
  3. Danışmanlar: PROF. DR. ABDULLAH HİLMİ LAV
  4. Tez Türü: Yüksek Lisans
  5. Konular: İnşaat Mühendisliği, Civil Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2025
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Lisansüstü Eğitim Enstitüsü
  11. Ana Bilim Dalı: İnşaat Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Ulaştırma Mühendisliği Bilim Dalı
  13. Sayfa Sayısı: 103

Özet

Karayolu taşımacılığının, bir ülkenin veya bölgenin ulaşım ağı içinde önemli bir payı vardır ve üstyapı yönetim sisteminin bu süreçteki etkisi oldukça büyüktür. Üstyapı yönetim sistemi karayolu taşımacılığında önemli bir rol oynamakla birlikte altyapı projelerinin daha verimli şekilde yönetilmesini amaç edinir. Etkin bir yönetim tartışmasız toplumsal fayda sağlamakla birlikte toplumsal ve ekonomik gelişmelerin de sağlıklı bir temele oturmasını amaç edinir. Üstyapı Yönetim Sisteminin kurulmasında en önemli kısım tartışmasız mevcut üstyapı performansının belirlenmesidir. Üstyapı yönetim sistemlerinin başarısı, bu ölçümlerin doğru bir şekilde yapılmasına ve elde edilen verilerin etkili bir şekilde analiz edilmesine dayanır. Günümüzde modern altyapı yönetim sistemlerinin verimliliğini ve etkinliğini artırmak için yapay zeka, büyük veri analitiği, makine öğrenimi, derin öğrenme ve akıllı algoritmalar gibi ileri teknolojiler kullanarak, üstyapı performans ölçümlerini daha doğru, hızlı ve verimli bir şekilde yapmak mümkün hale gelmiştir. Bu çalışmada yol yüzeyinin durumunu anlamak ve tahmin etmek için makine öğrenmesi yöntemlerinden sürekli bir çıktı değeri tahmin etmek için kullanılan denetimli öğrenme tekniği olan regresyon modeli kullanılmıştır. Uluslararası Pürüzlülük İndeksi (IRI) parametrelerini kullanarak yol pürüzlülüğünü tahmin etme üzerine odaklanan çalışma özel olarak donatılmış bir ölçüm aracı ile toplanan açık kaynaklı bir veri seti üzerinde yürütülmüştür. Veri seti, aracın kat ettiği mesafe ile birlikte sol ve sağ tekerlek yolları için IRI ve MPD ölçümlerini, ayrıca tekerlek yollarındaki çukur derinliklerini içermektedir. Yol yüzeyinin durumunu anlamak ve tahmin etmek için farklı regresyon modellerini karşılaştırmıştır. Bu modeller arasında lineer regresyon, ridge ve lasso regresyon, destek vektör regresyonu, rastgele orman ve gradyan artırma yöntemleri yer almaktadır. Her bir modelin performansı, Ortalama Kare Hata (MSE) ve Belirleme Katsayısı (R2) ile değerlendirilmiştir. Analiz sonuçları, rastgele orman ve gradyan artırma yöntemlerinin diğer modellere göre daha yüksek R2 değerleri ve daha düşük MSE değerleri ile daha iyi performans sergilediğini göstermiştir. Bu bulgular, yol yüzeyinin karmaşık özelliklerini modellemekte bu yöntemlerin üstünlüğünü işaret etmektedir.

Özet (Çeviri)

Transportation plays a critical role in the sustainable development of societies in economic, social and environmental areas. A safe, fast and economical transportation system provides effective distribution of goods and services, strengthens social ties, supports urbanization and economic growth. However, the effectiveness of the transportation infrastructure depends largely on the performance of the pavement. This performance directly affects the safety, comfort and economy of the transportation network. Road transport has an important share in the transportation network of a country or region, and the effect of the pavement management system in this process is quite large. The pavement management system plays an important role in road transport and aims to manage infrastructure projects more efficiently. An effective management provides unquestionable social benefits and also contributes to social and economic developments. The most important part of establishing a pavement management system is undoubtedly determining the current pavement performance. The success of pavement management systems is based on the correct performance of these measurements and the effective analysis of the obtained data. Nowadays, it has become possible to make pavement performance measurements more accurately, quickly and efficiently by using advanced technologies such as artificial intelligence, big data analytics, machine learning, deep learning and smart algorithms to increase the efficiency and effectiveness of modern infrastructure management systems. The use of machine learning technologies in pavement management systems plays a revolutionary role in increasing the safety and durability of road infrastructure. This technology can automatically detect and classify problems such as cracks, deformations and wear on road surfaces. Thus, maintenance needs can be predicted in advance and problems can be solved before they become bigger. The systems analyze various factors such as traffic density, weather conditions, and the properties of the materials used, and predict when and which sections of the road will need maintenance. These analyses, obtained from large data sets, make pavement management more economical and efficient compared to traditional methods. Thanks to machine learning algorithms, maintenance and repair processes are planned more intelligently, and resources are used at the right time and in the right place. This extends the life of the roads, increases driving safety, and significantly reduces costs. In short, these technologies make great contributions to society by optimizing both roads and resource management. In this study, regression model, which is a supervised learning technique used to predict a continuous output value from machine learning methods to understand and predict the condition of the road surface, was used. The study focused on estimating road roughness using International Roughness Index (IRI) parameters and was conducted on an open-source dataset collected with a specially equipped measurement vehicle. The dataset includes IRI and MPD measurements for the left and right wheel paths, along with the distance traveled by the vehicle, as well as the pothole depths in the wheel paths. Evaluating the pavement performance is a vital step in understanding the safety and durability of the infrastructure. Performance assessment examines physical and functional parameters to determine how well roads meet user needs. Among these parameters, the International Roughness Index (IRI) is one of the most critical indicators. The International Roughness Index (IRI) is an indispensable tool for evaluating the performance of transportation systems. By improving critical elements such as transportation safety, user comfort and economic sustainability, IRI ensures the establishment of a more effective and efficient transportation system for both individuals and societies. The open source dataset is LiRA-CD, which contains 1796 kilometers of road data collected from highways and urban roads in the Copenhagen area. In the LiRA-CD project, data collection was carried out through two main groups: electric vehicles and standard vehicles. Data collected from more than 50 sensors in the vehicle data of Electric Vehicles provides real-time information about road conditions.Through these sensors, detailed data are collected about road surface features, vehicle movements, steering wheel position, steering acceleration, steering angle offset, vertical acceleration, wheel torque, electric brake torque, engine revolutions per minute, engine slip torque, traction control torque, tire pressures, battery consumption, wipers and seat belt information, such as vehicle meta and movement characteristics, environmental factors and potential hazards on the road. In the data of Standard Vehicles, P79 Profilometer, ViaFriction measuring device and Automatic Road Analyzer (ARAN) were used for direct measurement of road surface. These vehicles measure standard parameters of road conditions such as International Roughness Index (IRI), Average Profile Depth (MPD) and wheel track depths. The CSV format datasets taken from the LiRA-CD project were first loaded with the read_csv function of the Pandas library, and different DataFrames were integrated into a single DataFrame using the concat function. This process ensures the integrity of the data processing and analysis processes. Then, the column names were arranged to prevent confusion in the analysis processes. The collected data was divided into 80% training and 20% test sets in order to test the generalizability of the model. This separation was made with the train_test_split function. It was trained using Python's scikit-learn library for model training and selection. The research compares various regression models to understand and predict the condition of the road surface. These models include linear regression, ridge and lasso regression, support vector regression, random forest, and gradient boosting methods. The performance of each model was evaluated using the regression performance evaluation metrics Mean Square Error (MSE) and Coefficient of Determination (R²). Mean Square Error (MSE) is related to the mean square error between the estimated and actual values, and Coefficient of Determination is one of the performance measurements with regression that allows curve fitting to be done more accurately in regression analysis. The criteria used to evaluate the performance of regression models measure how well the continuous values predicted by the model match the true values. Performance metrics for regression (regression related metrics or regression error metrics) typically involve calculating an error score to summarize the predictive skill of a model. The analysis reveals that random forest and gradient boosting methods outperform other models, demonstrating higher R² values and lower MSE, indicating their superiority in modeling complex characteristics of the road surface. The findings of this study contribute significantly to the development of road engineering and maintenance strategies. The results underscore the vital importance of accurately predicting road surface characteristics for road safety. Furthermore, the study highlights the value of utilizing such datasets for road maintenance planning and cost-effectiveness. Machine learning regression methods are an important tool for pavement performance measurements. These methods provide accurate predictions using big data analytics for efficient management of road infrastructure, timely maintenance and cost optimization. In this way, the safety, sustainability and economic efficiency of roads can be increased, road transportation and other infrastructure projects can be managed more economically, safely and environmentally friendly. These analytical approaches will continue to play a crucial role in future road maintenance and improvement efforts. In the future, prediction accuracy can be increased by collecting more data and using more sophisticated models. Furthermore, model performance can be further improved by using different sensor data and feature engineering methods. The generalization ability of the model can be increased by collecting more data; Model performance can be improved by applying other machine learning algorithms and hyperparameter optimization techniques for Model Optimization. The accuracy of the model can be increased by deriving new features and optimizing existing features.

Benzer Tezler

  1. Tez No

    255915

    The effects of geosynthetics on mitigation of rutting in flexible pavements

    Esnek üstyapılarda geosentetiklerin kullanımının tekerlek izi oluşumunun azaltımına etkisi

    İSMET ATALAY

    Yüksek Lisans

    İngilizce

    İngilizce

    2010

    İnşaat MühendisliğiBoğaziçi Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    PROF. DR. EROL GÜLER

  2. Tez No

    486542

    Bazik oksijen fırını cürufunun yapay agrega olarak esnek üstyapılarda alt temel ve temel tabakalarında kullanılabilirliğinin incelenmesi

    Study of the usability of the basic oxygen furnace slag in terms of artificial aggregate as lower foundation and foundation layers in flexible superstructures

    HİKMET ERGİN

    Yüksek Lisans

    Türkçe

    Türkçe

    2017

    Ulaşımİstanbul Teknik Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. ŞÜKRÜYE İYİNAM

  3. Tez No

    639770

    Mobilya endüstrisinde kullanılan plastik kökenli atıkların sıcak asfalt karışımda kullanılabilirliğinin araştırılması

    Investigation of the usability of plastic origin wastes used in the furniture industry in hot asphalt mixture

    YOUSIF RASHAD QADER

    Yüksek Lisans

    Türkçe

    Türkçe

    2020

    İnşaat MühendisliğiOndokuz Mayıs Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    DOÇ. DR. UFUK KIRBAŞ

  4. Tez No

    384948

    Backcalculation of pavement layer properties using artificial neural network based gravitational search algorithm

    Yapay sinir ağları tabanlı yerçekimsel arama algoritması kullanılarak esnek üst yapı katman özelliklerinin geri-hesaplanması

    ARDA ÖCAL

    Yüksek Lisans

    İngilizce

    İngilizce

    2014

    UlaşımOrta Doğu Teknik Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. ONUR PEKCAN

  5. Tez No

    856922

    Atık araç yağları ve atık bitkisel yağ katkılı bitüm içeren asfalt betonlarının performans özelliklerinin değerlendirilmesi

    Evaluation of the performance properties of asphalt concretes containing waste vehicle oils and waste vegetable oil additive bitumen

    BELİT KERMAN ERKARA

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Ulaşımİstanbul Teknik Üniversitesi

    İnşaat Mühendisliği Ana Bilim Dalı

    PROF. DR. ALİ OSMAN ATAHAN

    PROF. DR. MURAT KARACASU

Geri Dön